Look, is it serious that Claude consumes that much in terms of resources per token? Is that correct?
Even though it’s the best model, I don’t think it’s the most cost-effective one haha.
From the screenshot it looks like you used Opus with Max mode.
I will assume you mean Max mode as regular requests with Claude 4 Sonnet cost now less than 1 request.
Max mode charges by how much context (tokens) you send, it has to process and tokens it outputs. (thats how all AI API’s charge in reality).
So if you send a lot of context (many files etc) unnecessarily, it sends those files which are processed by AI as tokens. So yes it can add up a lot.
In Max mode it makes sense to focus the task on specifics instead of just attaching a lof files because context is large. It may be better to let the model explore what it needs by itself, since you are not charged per tool call as before.
Let me know what you tried and how, to get a better picture.
Several people reported to have had great success with Opus on very complex tasks, and thats where Opus is best at. Dont use it for simple tasks which Claude 4 Sonnet can handle. Same for any Max mode, its not best for simple tasks, but for cases where regular non-Max mode is not enough.
The Opus model only works in Max mode.